Adwatch

AdWatch is a comprehensive workflow solution designed for the automatic detection of audio and video content, specifically focusing on advertisements and promotional materials. This innovative system streamlines the process of identifying ads within media, ensuring efficient monitoring and management of advertising content.

Tech Stack

• Operating System: Linux
• Programming Language: Python
• Frameworks: Audfprint, FFMPEG, MoviePy

 

Dataset Requirements

• Image Dataset: A collection of advertisements and news domain images.
• Audio Dataset: .wav files of advertisements for database generation.
• Database: Contains audio fingerprints of advertisements stored in a pickle file.

 

Workflow for Ad Detection

1. Video to Audio Conversion: Convert video files (.mp4) to audio files (.wav) with standard sample rates and bitrates.
2. Ad Detection: Match audio fingerprints from the input file to detect advertisements.
3. Additional Frame Detection: Utilize Convolutional Neural Networks (CNN) for detecting additional frames related to ads.
4. Ad Removal: Remove detected ads from the original input video file.
5. Output Generation: Produce a final .mp4 file with advertisements removed.

 

End-to-End Development and Integration

1. Dataset Generation:
   • For Transfer Learning (VGG-16): Use .mp4 files of news and advertisements with a balanced time ratio.
   • Extract frames and annotate data according to the VGG-16 model requirements.
   • Convert .mp4 advertisement files to .wav format with a standard sample rate of 44.1 kHz and bitrate of 1411 kbps.
2. Training:
   • Fine-tune the VGG-16 model for binary classification of ads vs. news.
   • Generate a database of audio fingerprints using Audfprint.
3. Fusion of Audfprint and VGG-16:
   • Process input .mp4 files to extract audio and match ads using Audfprint.
   • Use CNN to predict ad frames, updating the time range for detected ads.
   • Remove ads based on the identified start and end times.
4. Testing:
   • Evaluate AdWatch by feeding in a video stream that contains news and advertisements.

 

Deployment Options

• Server/API: The system can be deployed as a server application or integrated into existing APIs for seamless operation.

 

Real-World Applications

• Audio and Video Content Detection: Automatically identify and classify audio and video content.
• Ads Verification: Ensure that advertisements aired as scheduled and verify compliance.
• Competitive Ads Analysis: Monitor and analyze competitor advertising strategies.
• Ads Affidavits Reports: Generate comprehensive reports detailing ad airings and compliance.

 

Specific Use Case

AdWatch is specifically designed to efficiently monitor media news campaigns, including news bulletins, talk shows, press conferences, and headlines. By removing news commercials that interrupt the flow of content, AdWatch enhances the viewing experience and ensures that key information remains accessible to audiences.AdWatch represents a significant advancement in automated media monitoring, providing users with the tools they need to manage and analyze advertising content effectively. Contact us today to learn how AdWatch can transform your media monitoring capabilities!